Low-passage primary human SC and A549 lung epithelial cells were cultured in DMEM/F-12 and DMEM media, respectively, as described previously (Chusri et al., 2016; Stone et al., 2019; Strange et al., 2021). ZIKV strain PRVABC59 (Human/2015/Puerto Rico) acquired from American Type Culture Collection was propagated once in Vero E6 cells.

SC and A549 cells cultured in 6-, 24-or 96-well plates were infected with ZIKV MOI of 1 for 1 h at 37°C as described previously (Siemann et al., 2017; Strange et al., 2021). In selected experiments, cells were treated with different recombinant proteins or dsRNA mimic poly I:C (InvivoGen) 24 h before infection. Recombinant human BMP6 (PeproTech) was used at 60 or 100 ng/mL concentration according to the instruction sheet, and 0.1% BSA was used as a control. Recombinant TGF-β1 and Activin A proteins were purchased from the R&D System. TGF-β1 was used at 10 or 15 ng/mL concentration while Activin A was used at 50 ng/mL or 100 ng/mL concentration with the respective vehicle as control. For activation of the innate immune pathways, poly I:C (InvivoGen) was used at the concentration of 10 μg/mL, and human recombinant IFN-β (R&D System) was used at 1 IU or 10 IU. After infection, the cells were replenished with fresh media containing different recombinant proteins or poly I:C in both infected and mock control cells every 24 h. Cells were harvested at 48 or 120 hpi.

SC grown to 60–70% confluency in a medium without penicillin–streptomycin were then transfected with 30 pmol of small interfering RNAs (siRNAs) using the Lipofectamine RNAiMax kit (Life Technologies). In brief, the siRNA-Lipofectamine RNAiMAX complexes (siRNA in 50 μl of Opti-MEM medium mixed with 3 μl of Lipofectamine RNAiMAX) were added to each well and mixed gently. After 24 h of incubation, half of the media was removed and replenished with fresh media. The silencing efficiency was quantified by RT-PCR and/or Western blotting at 48 hpt. The different pools of siRNAs (Silencer Select, Thermo Fisher) include siRNA for Toll-like receptor 3 (TLR3) (ID: s236), TLR7 (s27844), TLR9 (s28872), RIG-I (ID: s24143 and ID: s223616), and MDA5 (ID: s34498). A nontargeting siRNA was used as a negative control in every experiment (cat: 4404020).

ZIKV titers in cell culture supernatants at different time points after infection were analyzed by plaque assay using Vero cells and expressed as ZIKV PFU per mL of supernatant (Siemann et al., 2017). Intracellular viral RNA was extracted from cell lysates at 48 and 120 hpi and was measured by qRT-PCR using primer and probe specific for ZIKV Env region and expressed as PFU equivalents per μg of RNA as described previously (Siemann et al., 2017).

Changes in mRNA transcripts of key antiviral genes were measured from extracted RNA by qRT-PCR, as described previously (Strange et al., 2021). The reference gene GAPDH was used to normalize the data and calculate the fold-change of antiviral genes compared to respective mock-infected cells. Specific primer sequences used for GAPDH, IFNB1, IFIT1, MXA, STAT1, IFIH1 (MDA5), and DDX58 (RIG-I) transcript amplification have been previously described (Strange et al., 2018a). Forward and reverse primer sequences used for TGF-β family members, STAT1 and IRF7 are shown in Table 1.

Cell lysates were prepared using M-PER Mammalian Protein Extraction Reagent with a cocktail of protease and phosphatase inhibitors (Cell Signaling). The protein samples were separated on an NU-PAGE Bolt Bis-Tris Plus Gels (Invitrogen), transferred onto nitrocellulose membranes (Bio-Rad), blocked using Intercept (TBS) Blocking Buffer (Li-Cor Biosciences), and then incubated overnight with different antibodies (Strange et al., 2019). Specific primary monoclonal antibodies used were rabbit anti-human against RIG-I (Cell Signaling Cat. No: 3743S), MDA5 (Invitrogen Cat. No: 700360), MXA (Santa Cruz Biotechnology Cat. No: sc-271024), IRF3 and phospho-IRF3 (Cell Signaling Cat. No: 11904S and Cat. No: 4947S), STAT1 (Cell Signaling Cat. No: 9172S), phospho-STAT1 (Cell Signaling Cat. No: 9167L), STAT2 (Cell Signaling Cat. No: 72604S), phospho-STAT2 (Cell Signaling Cat. No 4441) and mouse anti-human β-actin (Sigma Cat. No. A2228-200UL), all at 1:1,000 dilution. Secondary antibodies (1:10,000 dilution) were conjugated with IRDye 800 and IRDye 680 (Li-Cor Biosciences), and blots were scanned using an Odyssey infrared imager.

Cell viability was measured in SC at different time points post-transfection and post-infection using the Promega CellTiter 96 AQueous One Solution Cell Proliferation Assay (Cat. No. G3582) as described previously and the absorbance was read at 490 nm on a Victor X Plate Reader (Strange et al., 2019).

ZIKV virus titers, cellular viability data, and gene expression mRNA fold-change are reported as mean ± SEM of at least three independent experiments. GraphPad Prism 5.0 (GraphPad software, San Diego, CA, United States) was used to conduct all statistical analyses. The unpaired Student’s t-test was used to compare differences between data from different groups. A value of p < 0.05 was considered statistically significant for all analyses.

Our previous studies collectively show that ZIKV titers in SC infected with both MOI 1 and 5 reach their peak at 48 h post-infection (hpi) and remain high till day 5 after infection (Siemann et al., 2017; Strange et al., 2019). Further, virus titers correlate well with robust antiviral response suggesting that the antiviral state in neighboring uninfected cells is established only after 48 hpi. Therefore, to evaluate the role of different PRR signaling in controlling ZIKV infection and spread in SCs we used the infectious dose of MOI 1 (Figure 1A) that did not exhibit any cytotoxicity till day 5 after infection (Figure 1B). We first validated our previous RNA Seq data by measuring the changes in the expression of different PRRs by qRT-PCR at 48 hpi. As seen in Figure 1C, ZIKV induced the mRNA expression of PRRs including MDA5, RIG-I, and TLR3 in SC that correlated with significant induction in IFN-β. The induction of MDA5 was most robust (40-fold), TLR3 and RIG-I were induced between 8 and 12 fold while TLR 7 and TLR9 were induced in the range of 2–4 fold. ZIKV-infected SC also showed activation of the transcription factor IRF3, as shown by the increased phosphorylation of IRF3, and increased expression of MXA protein, an indirect marker of IRF3 activation (Figure 1D). We next evaluated whether the downstream signaling of these PRRs is functional in these cells as these PRR pathways are not well characterized in human SC. SC were primed with poly I:C, a dsRNA mimic that is shown to activate TLR3, RIG-I, and MDA5 pathways, 24 h before infection. As expected, poly I:C treatment restricted ZIKV replication by more than 90% (Figure 1E). We further confirmed that the downstream IFN signaling in SC is comparable to other cell types. SC and A549 (lung epithelial cell-derived cell line routinely used to study ZIKV antiviral immunity) were treated with two doses of IFN-β before infection and we observed a significant reduction of ZIKV titers in a dose-dependent manner that was comparable in both cell types (Figure 1F). These data collectively show that viral RNA sensing pathways and IFN signaling are functional and are activated by ZIKV infection in SC.

TLR3 and TLR7 have been shown to sense ZIKV and other flavivirus RNA in various cell types (Hamel et al., 2015; Dang et al., 2016; Ojha et al., 2019). Based on RNA-Seq data that predicted activation of these TLRs, we first tested if they are critical in sensing and controlling ZIKV in SC. TLR3 and TLR7 were silenced in SC using specific siRNA and as seen in Figures 2A,B, the resulting mRNA expression levels of TLR3 and TLR7 were reduced by >80% at 48 h post-transfection (hpt) in comparison to controls without significantly compromising cell viability (Figure 2C). The cells were infected at 48 hpt and virus replication was measured. We observed no significant difference in the ZIKV titers in the supernatant at all the time points (24, 48, and 120 hpi) in siTLR3 transfected cells compared to controls (Figure 2D). We further analyzed if the silencing of TLR3 has any effect on the host response parameters. Consistent with the ZIKV replication, inhibition of TLR3 did not significantly affect mRNA expression of the antiviral genes IFNB1, MXA, and IFIT1 (Figure 2E). ZIKV titers were also measured in TLR7 silenced SC but no differences were found when compared to siCtrl cells at 24 hpi (data not shown) and 48 hpi (Figure 2F). Similarly, silencing TLR9 using siTLR9 (that reduced gene expression by >70% without affecting the viability, data not shown) also did not change the trend of ZIKV replication (Figure 2F). These results collectively suggest that TLR3, TLR7, and TLR9 do not play an essential role in antiviral response or restricting ZIKV infection in SC.

Based on the well-documented role of RIG-I in restricting ZIKV infection in other cell types, we next sought to determine whether RIG-I signaling plays a similar role in SC. We used two siRNAs targeting exon 3 and exon 11 of DDX58 to block RIG-I mRNA expression. Subsequent Western blotting analyses showed that the silencing efficiency at the protein level was 50 and 70% for exon 3 and 11 siRNAs, respectively (Figures 3A,B), with no significant change in viability at 48 hpt (data not shown). The difference at 24 hpi was not significant between the two groups, however, in contrast to TLRs, ZIKV titers increased in SC treated with siRIG-I exon 11 by almost 350% as compared to siCtrl treated cells at 48 hpi and the trend was also maintained at 120 hpi (Figure 3C). Since RIG-I is shown to control ZIKV replication in A459 cells, a lung epithelial cell line, routinely used to study ZIKV antiviral immunity (Lin et al., 2019), we also compared our data with A549 cells deficient in RIG-I. The percent increase in virus titers in SC was comparable to the increase in titers in A549 at 48 hpi (Figure 3D). We further compared the protein expression of the phosphorylated IRF3 (p-IRF3) using western blot, and as expected there was a marked increase in the p-IRF3 in siControl treated cells at 48hpi (Figures 3E,F). However, densitometric scanning analysis showed that although there was a trend of decrease in the p-IRF3 protein levels in SC silenced with siRIG-I, it was not statistically significant suggesting that silencing RIG-I has only a moderate effect in attenuating downstream signaling in these cells. These results collectively suggest that RIG-I is one of the PRRs involved in controlling ZIKV infection in SC.

Since RIG-I and MDA5 have nonredundant roles in detecting flavivirus infection (Errett et al., 2013) and because MDA5 induction was most robust in SC, we next tested the ability of MDA5 signaling to restrict ZIKV infection in SC. To attenuate the MDA5 pathway in SC, we treated the cells with siMDA5 and measured MDA5 protein expression at 48 hpt. As shown in Figure 4A, the expression of MDA5 was reduced by 60–70% at the protein level based on densitometric scanning analysis and the transfection did not compromise the cell viability (Figure 4B). To determine the effect of MDA5 silencing on ZIKV replication, SC were infected with ZIKV (MOI 1) at 48 hpt and virus titers were subsequently measured at 24, 48, and 120 hpi. Similar to what we observed with RIG-I silencing, MDA5 silencing also led to a significant increase (~200%) in ZIKV titers as compared to siCtrl cells only at 48 and 120 hpi (Figure 4C). We also compared the effect of MDA5 knockout on ZIKV replication in A549 cells. Interestingly, we observed that, unlike SC, loss of MDA5 in A549 did not lead to increased virus titers (Figure 4D), which was in agreement with other studies (Schilling et al., 2020) thus again supporting the cell-type specific role of different PRRs in virus-control. We further validated the association between silenced MDA5 and antiviral response and observed that there was a slight reduction in p-IRF3 levels in MDA5-silenced SC compared to control at 48 hpi (Figures 4E,F), however, it was not statistically significant. To further understand the synergy between the two RLRs, ZIKV titers were also determined in SC silenced with both RIG-I and MDA5 (Figure 4G). ZIKV titers in the double knockdown SC were significantly higher than the siCtrl cells at 48 hpi (Figure 4H). The percent increase of ZIKV titers in SC co-transfected with siRIG-I and siMDA5 was also significantly greater than the percent increase observed in individually silenced SCs (Figure 4I). Further, increased virus titers in the double knockdown cells correlated with decreased IFN response including mRNA expression of IFIT1 and MxA at the same time point (Figure 4J). These results collectively indicate that MDA5 is also one of the key PRRs in the response and restriction of ZIKV infection in SC, operating in parallel with RIG-I.

Our data in Figures 1–4 collectively show that although ZIKV is sensed by MDA5 and RIG-I and downstream antiviral signaling is activated, they were not capable of completely restricting ZIKV replication. Furthermore, we and others have shown that immunosuppressive pathways active in SC like Axl may modulate anti-ZIKV response (Siemann et al., 2017; Kumar et al., 2018). Therefore, we next evaluated if ZIKV infection alters the mRNA expression of TGF-β superfamily members, TGF-β1 and Activin A. As seen in Figure 5A, ZIKV does not induce changes in the mRNA expression of INHBA (Activin A) or TGF-β1. Since these molecules are also produced in high amounts by other testicular resident cells like LC, we next evaluated if exogenous exposure of TGF-β1 (10 and 15 ng/mL) or Activin A (50 and 100 ngmL) can affect ZIKV replication in SC. We found that exogenous treatment with recombinant TGF-β1 or Activin A at both lower (data not shown) and higher (Figure 5B) concentrations showed no significant change in ZIKV RNA. We further confirmed that the exposure of SC to these molecules also did not affect IFN-β levels (Figures 5C,D). The data suggest that the two TGF-β superfamily members TGF-β1 and Activin A do not respond to ZIKV infection of SC and do not play a role in modulating ZIKV replication and antiviral immunity.

BMP6, another member of the TGF-β superfamily, and its receptors are highly expressed in SC and regulate cell proliferation (Wang et al., 2017, p. 6; Ni et al., 2019). Further, ZIKV has been recently shown to induce BMP2 signaling in human brain pericytes (Chen et al., 2021). Therefore, we next assessed the gene expression of two key BMPs (BMP2 and BMP6) at different time points after infection. While ZIKV replication did not alter the levels of BMP2, we observed that BMP6 was modestly upregulated at 48 hpi by almost two-fold; however, by 120 hpi, BMP6 expression decreased significantly by >70% (Figure 6A). Additionally, the expression of the phosphorylated SMAD1/5, a marker of activated BMP6 signaling, decreased by 25–30% at 120 hpi as compared to mock controls, thus following the same trend of BMP6 mRNA expression (Figure 6B). Next, to assess whether BMP6 affects ZIKV replication, SC were treated with BMP6 at the concentration of 60 and 100 ng/mL based on other studies (Wang et al., 2017) for 24 h before infection and then replenished with fresh BMP6 every 24 h. We found that only 100 ng/mL BMP6 treatment resulted in an increase in viral titers at 48 hpi and a decrease by 50% at 120 hpi as compared to untreated SC (Figure 6C). We further examined the effect of BMP6 on the antiviral response and found that while the expression of key ISGs did not change drastically at 48 hpi (Figure 6D), the mRNA levels of MXA and IFIT1 increased significantly in BMP6 treated SC at 120 hpi by almost 3–4 fold (Figure 6E). Similarly, BMP6 treatment also induced mRNA expression of RIG-I and MDA5 at higher levels (~70%, Figure 6E) that correlated with decreased virus titers observed in these cells (Figure 6C). Together, these data suggest that BMP6 signaling is suppressed in SC at later stages of ZIKV infection and is one of the modulators of anti-ZIKV response in this cell type.

The increase in ISGs expression in BMP6-treated SC suggested that there might be crosstalk between the innate immune pathways and BMP6 signaling. To further validate the association between the antiviral response against ZIKV infection and BMP6 signaling, we first tested if BMP6 expression is under the control of IFN-β in SC and two other cell types (A549 and BMVEC). Our result showed no significant change in the expression of BMP6 in all three cell types at 24 h post-treatment of 1 IU/mL of IFN-β suggesting that BMP6 expression is not regulated by IFN-β (Figure 7A). Similarly, poly I:C treatment also did not affect BMP6 expression at 24 h post-treatment (data not shown) suggesting that BMP6 expression is not directly regulated by both molecules. We next examined whether BMP6 in turn can modulate IFN-β-associated antiviral response by treating SC with both BMP6 and IFN-β. Interestingly, treatment of BMP6 (100 ng/mL) increased the expression of ISGs including IFIT1, MXA, IRF7, and STAT1 (Figures 7B–D). To further evaluate BMP6 crosstalk with the antiviral response, we next treated SC with poly I:C in the presence and absence of BMP6. As seen in Figures 7E,F, BMP6 treated group exhibited significantly increased IFNB1 transcription compared to poly I:C treatment alone which also led to increased IFN-β levels in the supernatant media from cells at 24 h after treatment with poly I:C and BMP6 (Figure 7F). The increase in IFN-β correlated with a significant increase in the levels of IFIT1 and MXA transcripts (Figure 7G) in SC exposed to both poly I:C and BMP6. However, BMP6 treatment alone in the absence of poly I:C or IFN-β did not affect the basal expression of these antiviral genes compared to untreated controls (data not shown). To further assess the involvement of downstream transcription factors in BMP6 crosstalk, we measured the levels of p-IRF3, a critical transcription factor in RLR signaling, in SC after 24 h of treatment with IFN-β and poly I:C in the presence and absence of BMP6. While the levels of p-IRF3 were very low and not different between the untreated and BMP6 exposed cells, the levels of p-IRF3 markedly increased in both IFN-β and poly I:C treated cells. However, the presence of BMP6 further increased p-IRF3 levels in both poly I:C and IFN-β treated cells by 40 and 60%, respectively, (Figure 7H). Analysis of transcription factors involved in IFN signaling also demonstrated a significant increase in p-STAT1 levels in SC treated with both BMP6 and poly I:C and IFN-β by 100 and 350%, respectively. However, there was no difference between p-STAT2 levels in BMP6 treated and untreated groups (Figure 7H). Overall, these data showed that although type I IFN does not directly affect the BMP6 expression, BMP6 signaling is involved in positively regulating antiviral response in SC.